JP2003021630A - Method of providing clinical diagnosing service - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method of providing the clinical diagnosing service. SOLUTION: The method of providing the clinical diagnosing service is provided. This method includes a stage for collecting biological samples, a stage for analyzing the biological sample to determine a composition of a genetic material of the biological sample, its behavior or at least a part of protein, a stage for reporting a result of the analysis (for example, to a medical care provider), and a stage for taking the information obtained by the analysis, in the analysis of the follow-on biological sample. The information obtained from the analysis can be taken in the follow-on analysis, for example, by using the information to improve the algorithm of the used information product or a database component, or can be used to improve the reliability in statistical analysis. The database system and a device to execute this method, are similarly provided.

Description

Detailed Description of the Invention

BACKGROUND OF THE INVENTION The present invention relates to the fields of clinical diagnostics and laboratory medicine.

Genetically based diagnostics are rapidly becoming the standard tool in clinical laboratories. These diagnoses attempt to correlate one aspect of the genetic composition or behavior of genetic material in vivo with a physiological condition, disease state or propensity for disease. This includes analysis based on the presence or absence of gene mutations such as sequence insertions, deletions or inconsistencies. Similarly, it also provides information about how gene expression occurs within an individual or a portion thereof (eg, a cell), such as whether certain expression is upregulated or downregulated. May be included.

The usefulness of diagnostic methods depends on the output of the bioinformatics system used to perform the above correlations. Most of these bio-scientific systems arrange sequences (nucleotide bases or amino acids) in a defined format for the user
Are required to submit. The system then algorithmically engages in comparing sequences to other known sequences or to comparing gene expression profiles to other expression patterns. Then, the similarity between the known sequence and profile and the sequence and profile of the sample is
Compared or "rated" according to various rules. Similar to a known sequence in the system if the sequence compared to the unknown sample is known to have some physiological effect or be representative of one condition or disease state An unknown specimen can be said to have the condition or disease state. Bioinformatics systems that use algorithms to analyze sequence similarity include B
Included are the LAST and FASTA computer programs. The robustness of databases used to compare genetic information from unknown specimens with genetic information that reflects known symptoms is important.

The algorithmic aspects of bioinformatics systems likewise affect the usefulness of diagnostics. The programming logic and statistical and mathematical relationships used to determine when one array is similar to another are:
Central to the usefulness of these systems as an aid in making diagnostics and prognosis. However, there are more fundamental biological components for bioinformatics. That is, the functionality is attributed to sequence identity and expression.
If the exact relationship between the symptoms of the problem and the genetic information were known, this would not be a troublesome problem. Of course, this is not the case here. Some diseases or conditions have been found to directly correlate with certain genetic profiles, but most are completely unknown or poorly known. The probability of being able to properly assess a disease state or symptom improves as more factors are identified in the genetic profile associated with these symptoms. For example, although p53 mutations are a frequent event in some cancers such as colorectal cancer, so far any specific p53 mutation or p53 mutation group has determined colorectal cancer. Cannot be used to make a diagnostic diagnosis. P53 as a marker for colorectal cancer, Asco Online, http://www.asco.org/prof/pp/html/m-tumor8.htm
See. Some have speculated that epigenetic changes such as DNA methylation may also have diagnostic or prognostic value associated with colorectal cancer. Pharoash and
Caldas, "Molecular Genetics and Human Cancer Assessment," Expert Re
views in Molecular Medicine, http: //www-ermm.cbcu.
See cam.ac.uk/99000526h.htm. And, in addition,
One might speculate that the presence of both p53 mutations and DNA methylation at certain sites would improve the probability of accurately diagnosing colorectal cancer. As additional profile elements are identified, the databases and algorithms used to compare normal and diseased or affected genetic material must be updated to realize these improvements.

Diagnostic services are usually provided by the laboratory at the direction or request of a healthcare provider. The laboratory receives a sample of the patient from the healthcare provider, then performs diagnostic tests, obtains the results, and then communicates these results to the patient or the healthcare provider. This model also applies to gene-based diagnostics, such as those that rely on amplification of genetic material. As mentioned above, the analysis of the results of genetic-based tests involves robust database algorithm operations. These algorithms can be updated regularly as new information about the genetic profile becomes available, which requires waiting for clinical information to be pursued and incorporated into such information products. Thus, the process can be divided into two at best. In one aspect of the standard process, a patient's genetic material is analyzed. In an entirely different aspect of the process, the information products used in the analysis are newly created,
Provided to the parties performing the analysis. In such a process, there is no way to continuously improve the robustness of the database, the power of the algorithms used to perform the analysis, and the confidence interval of the results obtained from the process.

Artificial Neural Networks (ANNs) have been proposed as one way to create new powerful algorithms for processing diagnostic information.
US Pat. No. 6,058,322 to Nishikawa and Ba
U.S. Pat. No. 5,769,074 to Rnhill is an example. ANN does not solve existing problems.

ANN as described by Barnhill
Compares various data with a trained network to see the significance attributed to each data component. For example, if one sample was being analyzed to diagnose prostate cancer, PSA and age may be the two data elements that the network is trained to take into account. This network is 1
For a given PSA concentration at one age,
It could also be trained in such a way that it can be given greater weight as an indicator of the presence of cancer compared to the same PSA level at different ages.

These ANNs solve multivariate problems by forming multivariate (weight) mathematical models based on examples and then fitting them to realistic cases. This process is commonly referred to as "training." The network itself can ultimately choose the best rule to use for data comparison. However,
The ANN must be trained in such a way that it meets specified statistical requirements (eg confidence level and positive predictive value) before it can be used. In this sense
An ANN, such as the one described in the Barnhill patent, is static. There are discrete uses of data such as training, tests or sample cases. Training is not a continuous process.

Another notable feature of the Barnhill patent is that
The point is that the comparisons it makes are necessarily based on "normal" values achieved through statistical analysis as part of the training process. The act of training is itself
It is the act of determining or setting the normal range. Once trained, the ANN is queried to compare these normal values with actual patient data to assess diagnosis or prognosis. Beyond the algorithmic aspects of ANN, this is rather a standard treatment of the data for clinical measurements of standard serum markers such as PSA. In the absence of an ANN, the physician would simply compare the level of the marker with the normal value provided in the reference. ANN
The power of is to enable us to construct a normal range in a way that accounts for many variables that are difficult for humans to consider at the same time.

[0010] Any ANN has a genetically relevant indicator (eg specific deletion sequence, epigenetic mutation) to improve the genetic or clinical profile of diagnostic algorithms and databases in relation to their diagnostic or prognostic relationship. )
Does not propose a process to expand or contract the number and / or significance of

US Pat. No. 6,056,6 to Roberts
No. 90 is Bayesi in building a diagnostic decision support tool
I suggest using an network. Bayesian networks, also called credit networks or causal stochastic networks, use probability theory as the basis for analogy under uncertainty. Baye explains the analogy
The ability of the sian network is one important excellence over most ANNs. Despite this, Ro
berts does not propose to improve the analogy process itself as a function of clinical use of the system.

US Pat. No. 5,966,71 to Adams
No. 1 proposes to use an autonomous intelligence agent to update the database and algorithms from the result table. The patent is directed to the structure of algorithms and database systems that interact with each other. In this system, the updated components can communicate with the base system when the base system needs help, such as when the sequence search reveals no tight matches at all. The patent also does not address the validation of the data used to form the daemon update program, nor does it address the source of the data. Without validation, operations that constantly try to improve statistical reliability based on increasing sample size can run into problems. For example, if the daemon program contained gene expression data that was not in the base system and was not validated, its use would in fact increase the uncertainty of the results produced. Moreover, the patent does not even show that improved statistical reliability is possible. This is because the daemon is used to intervene only information and programming steps that did not previously exist in the base system. There is no mention of using such daemons to reintroduce existing information and increase the sample size from which statistical belief is achieved.

US Pat. No. 5,024,699 proposes the establishment of a system for entering patient test results and providing clinical advice to the patient based thereon. The patent describes a process by which the drug dose algorithm is modified based on these results.
The algorithm in this case is only relevant to the target patient for whom the result was entered. It is not a systematic algorithm that affects the way data is interpreted across the patient pool.

It would be beneficial to have a method of providing these analytical and diagnostic services that continuously upgrades the power and usefulness of the information products used in providing the analytical and diagnostic services. The ability to combine diagnostic information from various sources will improve the accuracy and accuracy of genetically based diagnostics. By delivering diagnostic services by distributing the tasks involved, the efficiency, timing and quality of the services performed will also be improved.

SUMMARY OF THE INVENTION The present invention comprises the step of analyzing the results obtained from the testing of biological specimens to determine the composition of its genetic material, its behavior or at least part of the protein in a method for providing a clinical diagnostic service. , And incorporating the information obtained through the analysis into a subsequent analysis of the biological specimen.

[0016] Another aspect of the invention is a method for providing a clinical diagnostic service, wherein the biological specimen is used for determining the stage of collecting the biological specimen, the composition of its genetic material, its behavior or at least part of its protein. The method comprises the steps of: analyzing, analyzing the results of the analysis (eg, to a healthcare provider), and incorporating the information obtained through the analysis into a subsequent analysis of the biological specimen. The information obtained from an analysis can be incorporated into a subsequent analysis, for example by using it to improve the algorithm or database component of the information product used, or otherwise statistically. It can also be used to improve reliability.

The present invention also provides systems for utilizing the above-described methods and products useful in such systems (eg, computer-readable media containing instructions for executing algorithms and manipulating databases). Also included.

DETAILED DESCRIPTION Definitions: The following terms are used throughout the specification.

By "internal database" is meant a database containing biomolecular sequences (eg nucleotides and amino acids) with which sample sequences or profiles are compared. This includes information associated with a sequence, such as the library in which a given sequence was discovered, the manifestation of physiological symptoms associated with that sequence, and the association of genetic material behavior or sample sequences with symptoms or disease states. It may contain any other information that will help you. In addition, the database may be derived from gene expression patterns characteristic of cell or tissue type, DNA methylation patterns characteristic of cell or tissue type, or of any other hereditary or somatic origin characteristic of cell or tissue type. It may include genetic variation. The internal database directs information about the biomolecule sequences that are embedded data structures, or utilizes sequence database components found in separate discrete databases that are accessed by the internal database as needed. .

An "analytical database" is a class of internal database used as a reference in the process of determining some information about a cell or tissue that requires characterization. For example, it may be possible to determine whether a cell or tissue removed from a patient exhibits characteristics of the cell or tissue that require some form of medical intervention that may be beneficial to the host of this cell or tissue. May be advantageous. This kind of analysis can be described as screening, diagnosis, prognosis, or else it can be a surveillance procedure. The main feature of any analytical database is that the data it contains is capable of comparing the subject's information to criteria for which the feature has already been characterized, and makes a conclusion about the test substance at a predetermined confidence level. It is at least partially organized in such a way that

A "discovery database" is a class of internal database that contains sequence or pattern data collected from a wide variety of sources. The discovery database is analyzed to identify sequences or patterns that may be useful as a component of the analysis database. Once one component of the discovery database has reached a defined significance level, it is placed in the analysis database. This can be done according to pre-programmed rules. The discovery database has an ordering level that allows multiple queries with multiple parameters simultaneously or sequentially. Typically, the data entered in the discovery database will include genetic data annotated with clinical information. This reflects the currently acceptable circumstances of protecting patient privacy. For example, one entry into the database could be the RNA expression profile of a biopsy from a suspicious prostate cancer in which the expression data is electronically linked to the patient's history and a complete profile of the current disease state. Mechanisms can be used such that subsequent data about the patient is collected and added to the annotation field for that pattern. The data describing the patient will be anonymous or encoded, and the entries in the database can be encoded (eg, using the tags described below in different contexts). The code is sent either to the patient or the doctor and at the time of presentation the new data is sent attached to one code. The code allows proper containment of annotations. Only the person with the code, the doctor or the patient,
Provides access to identifiable data (relative to the patient).

A "reference pattern" or "reference sequence" is a sequence or pattern that has been identified in a discovery database and shown to have diagnostic or prognostic utility. The reference sequence or pattern is typically found in a discovery database and then exported into an analytical database for use in medical practice. The flow of reference material is usually one-way from discovery to the analytical database, while the sequence or pattern flow that must be determined in the future as all or part of the reference sequence or pattern begins with entry into the analytical database and ends with discovery. The export to the database can be followed, otherwise it may be entered directly in the discovery database.

"External database" means a database located outside the internal database. Typically this is maintained by a different company than the one that maintains the internal database. In the context of the present invention, the external database is used first of all to obtain information about the various sequences stored in the internal database. External databases can be used to provide some descriptive information stored within gene expression databases. In a preferred embodiment, the external database is GenBank and its associated databases maintained by the National Center for Biotechnology Information (NCBI), which is part of the National Library of Medicine. Gen Pept is a companion public protein sequence database that includes all protein databases from Gen Bank. Other examples of external databases are the Blocks database maintained by the Fred Hutchinson Gunn Research Center in Seattle and the Swiss Prote site maintained by the University of Geneva.

"Record" means an entry in a database table. Each record contains one or more fields or attributes. A given record may be uniquely identified by one or a combination of fields or attributes known as the record's primary key.

The term "sequence" in the case of nucleic acids refers to
By such nucleotides is meant comprising the nucleic acids in the order in which the nucleotide or nucleotides constitute the nucleic acid. In the case of a protein, it means the amino acid or amino acids that comprise the protein in its constitutional order.

"Pattern" refers to known and sample genetic material or protein structure (eg, amino acid sequence).
By sequence or group of sequences that form the basis of the comparison between. The pattern can be the behavior of gene sequence groups. For example, the pattern can be the relative gene expression activity of a set of defined genes whose observed behavior is characteristic of a particular physiological activity such as apoptosis or diagnostic or development of one disease. . Moreover, the pattern of relative gene expression levels may be indicative of possible developmental processes of cancer cells or tissues. This type of pattern is sometimes referred to as a cell or tumor profile, genetic signature or expression profile. Therefore, the act of pattern determination is commonly referred to as profiling. In addition, the pattern may include other structural or behavioral identifying features of the genetic material, such as epigenetic modifications. For example, the pattern can be the status of DNA methylation of a group of genes. The methylation pattern can be the relative hyper- or hypomethylation status of a number of genes, the methylation pattern being characteristic of a particular physiological activity or diagnosis, such as apoptosis, or a characteristic of the development of one disease. obtain. Moreover, the pattern of DNA methylation may be indicative of possible developmental processes of cancer cells or tissues.
The pattern can also be a group of genetic alterations or mutations such as a single nucleotide polymorphism (SNP) group. For example, if the SNPs are reproducible in view of their coexistence within an individual's genome, and there is conviction that these SNP groups are correlative and / or prophetic, then these SNPs will have one pattern Make up. The SNP pattern can include SNPs that are spaced throughout the genome, or the SNP pattern can form a haplotype in which the co-inherited SNPs are in linkage disequilibrium. Patterns can also include conserved concomitant events that can be derived from any of the genetic events described above, eg, one pattern is SNP within a specific gene, of 20 distinct genes. It can include specific relative expression levels, reproducible deletions of chromosomal deletions (eg in loss of heterozygosity) and distinct chromosomal hypermethylated regions. The defining feature of this collection of events into a pattern is that they are predictive, diagnostic or prognostic of overt phenotypes or diseases within the same individual that carry all of the genetic alterations.

The "behavior" of genetic material means the unambiguous outline of a sequence. In the case of nucleic acid sequences, the expression of a gene or sequence is one measure of the behavior of that sequence.

Sequence Analysis Methods for determining nucleic acid sequences are now well known. Primary nucleotide sequencing can be complemented by any number of methods, including dideoxy terminator sequencing. Analysis of the presence or absence or relative quantification of relative levels of RNA or DNA showed that Northern, Southern blotting, in sit
u can be complemented by a number of published methods such as hybridization, slot or dot blotting. More recently, microarray technology has been used to determine whether various sequences are present and whether the identified genes are expressed.
Several examples of such microarray technology are provided in US Pat. No. 6,004,75, each incorporated herein by reference.
5; 6,051,380; 5,837,832. These methods utilize a substrate to which is attached a number of standardly labeled oligonucleotides. When a sample containing a sequence that is complementary to the bound oligonucleotide is contacted with the substrate-bound oligonucleotide, the method utilizes some form of signal to indicate that hybridization has occurred. For example, solution-based molecules, typically specimens, can be labeled and the presence of the label can be detected by fluorescence microscopy or radiography. Alternatively,
The two molecules combine to give rise to some detectable phenomenon such as fluorescence. Microarray-based methods can take advantage of a number of different technologies (eg, some passive and some active), all of which identify and characterize a fixed number of sequences simultaneously. Have the potential to attach. Using other methods to analyze juxtaposed numbers of sequences, including cDNA sequencing, serial analysis of gene expression (SAGE) and use of solution-based arrays in which specific oligonucleotides are linked to tagged lymphocytes You can also do it. Following solution hybridization, hybridization activity is detected by a range of published methods. Although any method for determining nucleic acid sequence can be used in conjunction with the practice of the present invention, the described highly parallel methods such as the microarray approach are most preferred. Methods for determining amino acid sequences are also well known.

To practice the method of the present invention, sequence information or gene expression profile is obtained. Therefore, at some point in time, a sample of the patient must be obtained. There is no limit to the type of sample that can be used, provided that the sample can be assayed to determine sequence information. Thus, specimens can be obtained from circulating blood, tissue biopsies, washes, and any other method that will capture sequences. A complete set of methods for extracting such specimens is available.

Sequence information can be generated and rendered in a variety of ways. For example, if a microarray with bound fluorescently labeled oligonucleotides is used, a reader can be used to generate a graphic representation of each bound sample oligonucleotide. These graphics can be digitized in such a way that the intensity of each detectable event can be measured. This can be very useful in gene expression analysis, where the determination of RNA segment production is an important indicator. Alternatively, one or more PCs may be used to simply indicate whether a particular segment is present.
It is also possible to use the R reaction. Information at this time,
It can be arranged in the form of a table, database, etc.

Any method of displaying sequence information or gene expression profile can be used in the practice of the present invention.

Bioinformatics As noted above, much of the diagnostic utility of bioinformatics systems stems from the process of comparing or matching sample sequences or expression patterns to known sequences or known expression patterns. ing. Various techniques are available for this purpose. Since pattern matching between known patterns and sample patterns is performed, structural data (eg genomic sequences) can be taken using the same or similar approaches.
And expression data (eg, gene expression profile) can be compared. Query array (array of array list)
Using the nucleotide sequences from patient specimens as, databases containing previously identified sequences can be searched for regions of homology. Examples of such databases include GenBank and EMBL.

One homology search program that can be used is a paper by DJ Lipman and WR Pearson entitled "Fast and Sensitive Protein Similarity Search" (Science,
227, 1435 (1985)). In this algorithm, regions of homology are searched in two stages. In the first stage, the region of highest homology is determined by calculating the match score using the homology score table. At this stage, 2
The parameter "Ktup" is used to set the minimum window size that should be shifted for comparing two arrays. Ktup also sets the number of bases that must match to extract the region of highest homology from the sequence. At this stage, no insertions or deletions were applied and the homology was early (INI
T) is displayed as a value. In the second step, the regions of homology are aligned to obtain the best match score by inserting gaps to add putative deletions. The match score obtained in the first stage is recalculated using the homology score table and the insert score table up to the optimized (OTP) value in the final output.

The DNA homology between two sequences is determined by the method of constructing a dot matrix homology plot by Harr (Needleman,
SB and Wunsch, C, O., J Mol. Biol 48: 443 (1
970)) can be used to graphically inspect. This method produces a two-dimensional plot that may be useful in determining the relationship between regions of homology and repeat regions.

However, in a preferred embodiment class, the comparison between the nucleic acid sequence and expression data obtained from the specimen and the reference pattern is performed using Factura software (also
Applied Biosystems Inc, including software known as Applied Biosystems Inc.)
INHERIT available from (Foster City. Calif)
It is accomplished by processing the data obtained from patient specimens in a commercially available computer program known as the 670 Sequence Analysis System. The Factura program is po
Each sample sequence is pre-processed in order to "edit out" those parts that do not appear to be advantageous, such as the ly A tail and repetitive GAG and CCC sequences. Low-end search programs can be written to mask out such "low information" arrays, otherwise BL
Programs such as AST can ignore low information sequences.

In the algorithm implemented by the INHERIT 670 Sequence Analysis System, a pattern specification language (developed by TRW Inc) is used to determine regions of homology.
"There are three parameters that determine how INHERIT analysis performs sequence comparisons: size,
Window offset and error tolerance. The window size specifies the length of the segment into which the query array is subdivided. The window offset identifies where to start the next segment [to be compared], counting from the beginning of the preceding segment. The error tolerance specifies the total number of insertions, deletions and / or substitutions allowed over a particular word length. The error tolerance can be set to any integer between 0 and 6. The default settings are window tolerance = 20, window offset = 10, and error tolerance = 3. " INHERIT Analysis User Manual P2-15, Version 1.0, Applied
Biosystems, Inc. October 1991. A combination of these three parameters can be used to search the database for sequences containing regions of homology, with appropriate sequences scored at the initial value. Then, these regions of homology
It is examined using a dot matrix homology plot to determine the relationship between homology and repeat regions. Smith-Wa
It is possible to display the results of homology searches using terman alignment. INHERIT software may be executed by a Sun computer system programmed with the UNIX® operating system.

As an alternative to the search for INHERIT, the BLAST program, GCG (Genetics Compu
ter Group WI) and Dasker program (Temp
le sample ith, Boston University, Boston, MA) is included. Nucleotide sequences are GenBank, EMBL or GE
It can be searched against custom built internal databases such as NESEQ (available from Intelligentics, Mountain View, CA) or other internal databases for genes.

BLAST (Basic Local Alignment Sear
ch Tool) program and the Smith-Waterman algorithm look for regions of similarity without gaps between two sequences.
To do this, they determine (1) the alignment between similar regions of two sequences and (2) the percent identity between sequences. Alignment is calculated by aligning regions of substantial similarity on a base-by-base basis. In these regions, identical bases are scored with a value of +5 and mismatched bases are rated with a value of -4 (for nucleic acids). Regions of adjacent bases with sufficiently high scores are considered high-scoring pairs ("HSPs"). In BLAST, the highest HSP score (called the BLAST score) is presented as one output. Furthermore, each H
For SP, the percent identity is calculated and presented as the BLAST output as well as the alignment. Finally, the P-value for each HSP is calculated. P-values represent the probability that the observed similarity was a result of random appearance. Lower P-values represent greater confidence that the observed similarities are not due to random events.

The product score represents a normalized summary of the BLAST output parameters and is used to express the value of the alignment between a query and the matching sequences. Specifically, the product rating is BLAST
It is a normalized value in between, which represents the strength of the match.
This represents a balance between fractional overlap and BLAST alignment quality.

Many other sequence matching / analysis algorithms are available. For example, the FASTA method first compares the maximum number of short perfect sequence matches in a process called hashing. The sequence showing the highest match is then subjected to a second analysis, which assesses the match according to criteria other than the one used in the first comparison. Finally, the sequence with the highest match is aligned and a score is given based on the parameters for alignment proximity.

In one aspect of the invention, the matching algorithm and associated database may comprise part of a system used to arrive at the diagnosis, prognosis or staging of a condition or disease state. Another aspect of the system is an internal database that is continuously updated so that the sequence assessed during the analysis of each sample is incorporated into the analytical database used to compare the subsequent sample sequences. Is. That is, the sequences generated from the patient sample analysis are later incorporated into the reference pattern.

The databases used to match known sequences or profiles with the patient's specimen nucleic acid sequences or gene expression profiles also result from sequences that have been identified for clinical significance by correlating those sequences with clinical outcome. I think. These correlations can be stored in and manipulated from the same database used to determine homology, or otherwise in another database that interfaces with the homology determining database and algorithm. May be stored and maintained in. In one example, a nucleic acid sequence that represents amplification of the her-2-neu gene, along with the presence or absence of other undiscovered nucleic acid sequences, may indicate that the patient is developing invasive breast cancer. There is. Similarly, increased or significantly reduced expression of a gene may also represent uncontrolled growth of one cell type. Once homology or pattern similarity between these sequences or gene expression profiles and those of patient specimens has been established, the sequences or profiles are clinically attributed to them in the analytical database. Aligned with meaning. At this time, her-
In the case of the 2-neu gene, clinical results (ie, information) are generated indicating that the patient is developing aggressive breast cancer.

The establishment of the gene expression profile is performed through the following process, which will be useful in predicting whether a previously identified patient with a tumor will relapse in the future. Create a class prediction model: (1) define discriminatory relationships (eg, recurrence vs. survival); (2) rate individual genes for their ability to predict desired patterns and the statistical significance of these scores. (3) selecting a subset of informative genes;
(4) Build prediction rules based on subsets; (5) Validate rules on initial data set and independent data. Such schemes have been successful in analyzing data from a wide range of tumors. These methods typically vary with respect to score selection, significance calculation and exact rule construction methods.

In order to select for a particular gene expression marker, each gene on the microarray of genes displaying or associated with cancer is labeled with a "similarity" of each such gene, with two distinct classes of desired discrimination. It is rated according to "sex". Different distances and measures can be used as such scores. From this process, gene lists are generated and further narrowed according to additional considerations for the purpose of generating signature subsets.

Predictors are constructed from the narrowed signature subset list. Within the predictor, each gene casts a weighted vote for one of the classes (recurrence or survival) and gained more votes (above a given given victory margin) Wins that prediction. The weight of the vote for each gene depends on its expression level in the new specimen and its "quality" as reflected by its score. The votes for each class are summed up,
It is compared to determine the prediction strength, which is a measure of the winning class as well as the winning margin. A sample is assigned to the winning class only if the predicted strength exceeds a given preset threshold.

For the example used in the demonstration, the predictor is preferably combined with an independent data set, since most classification methods will work.
Mutual validation is checked and evaluated. The sample may be divided into more than one group for validation. Otherwise, commonly used cross validation methods such as leave-one-out cross validation (LOOCV) may be used. Multivariate analysis can then be applied to test the link between patient prognosis data and assessed marker expression.

An example method for comparing expression information is as follows: hybridizing labeled cDNA molecules to a microarray containing complementary nucleic acid sequences and one label (eg, using a fluorophore). Let it form. The microarray is then scanned and the intensity of the spots recorded. Then, a matrix of intensity data is created.

After that, a reference gene expression vector is created. If A, B ... Z are used to represent the sample groups to be differentiated, then a, B is used to represent the number of samples used in each group to construct the reference gene, respectively.
b, ... Z are used. Therefore, the notation A ₂₁ represents the expression intensity from the second gene in Sample 1 of Group A. If each specimen was hybridized on a microarray with genes of size n, then the matrix A, B,
... Z represent expression data from all of groups A, B, ... Z, respectively.

[0049]

[Equation 1]

The geometric mean expression value for each gene in each matrix is then calculated in such a way that the following matrix is created (A ₁ (geomean) is set {A ₁₁ A ₁₂ ... A
_1a } is the geometric mean of genes in group A 1),

[0051]

[Equation 2]

The reference gene expression vector is the geometric mean of these vectors.

[0053]

[Equation 3]

In the formula

[0055]

[Equation 4]

Is {A _{1 (geomean)} B _{1 (geomean)} ... Z
It is the geometric mean of _{1 (geomean)} .

After the reference gene expression vector is created,
The original data set is transformed by taking the log of the ratio in relation to the reference gene expression value for each gene. In this way, the matrix {A′B ′ ... Z ′} is generated.

[0058]

[Equation 5]

Here,

[0060]

[Equation 6]

It is At this time, the value represents the fold increase / decrease with respect to the average for each gene.

At this time, genes having weak differentiation power are removed from the matrix {A'B '... Z'}. 1 to n
For the genes i up to, the values {A ′ _i1 , A ′ _i2 ... A ′ _ia , B ′ _i1 , B ′ _i2 ... B ′ _ib , expressed as an anonymous number,
If none of Z _it , Z _it ... Z _i2 } is greater than the threshold (In3 in the preferred embodiment), then gene i is removed from all matrices. In other words, this value must have at least one value in every matrix with an absolute value above the threshold (preferably In3) in order to be considered a diagnostically relevant gene. The matrix from which the genes having only weak differentiation power are removed is the matrix {A ″ B ″ ... Z ″}.

At this time, the signature extraction algorithm is applied to each of the resulting matrices {A ″, B ″ ... Z ″} in order to newly create a signature as follows. Is M
Called axcor algorithm, each group {A ″, B ″
, Z ″} acts separately. For each column pair in the matrix, genes that are coordinately higher, mean and lower than the mean value (defined below) are 1, 0, and −, respectively. Given a value of 1, each produces a weight vector representing that pair. For the matrix A ″

[0064]

[Equation 7]

Calculations are performed in pairs of two. The final mean weight vector, called the signature for group A, is

[0066]

[Equation 8]

It is calculated by taking the average of the weight vectors. Thus, the signature contains the same number of genes as A ″, the values of which are −1 and 1 respectively representing genes that were consistently expressed at low and high levels in relation to the mean of all groups. , [-1, 1]
Should be inside.

In the above-mentioned two-pair calculation, the coordinate sequences c1 and c2 are taken and cli

[0069]

[Equation 9]

(In the formula

[0071]

[Equation 10]

Is the mean of column c1 and Sc1 is the standard deviation). For each gene pair in c1 'and c2',
The products are then stored in vector p12 and each value in p12 is then sorted from lowest to highest. The nominal cutoff value (0.5 in the preferred embodiment) is then used to collect all genes with larger product values in p12. The Pearson correlation coefficient for this gene set is then calculated using the values in columns c1 and c2. The cutoff value is then increased until the correlation coefficient is greater than a statistically relevant number (0.8 in the preferred embodiment). When this is complete, the gene sets that meet this criterion will have cl 'and C
A 1 is assigned if both gene values in 2'are positive, and a -1 is assigned if both gene values are negative. c
0 is assigned for all other genes in 1'and c2 '. The resulting vector is a weighted vector that represents the pair. Values of -1 and 1 represent genes that were consistently expressed at low or high levels respectively in relation to the mean of all groups.

Once the signature has been created, the unknown specimen can then be evaluated against it. Prior to the rating, the genes in the sample S with weak differentiation values are removed in such a way that the remaining rows are the same as those in the signature vector, thus creating a new sample vector S ″. , S ″ for each gene and its weight in the signature vector. For example, the scores between the sample vector S "and signaling Nichua vector A ^s

[0074]

[Equation 11]

It is The normalized score is the standard deviation of (score-randomized score) / randomized score, where the randomized score is the score between S ″ and the signature vector whose gene position was randomized. There is typically 10 to calculate the mean and standard deviation.
Zero randomized scores are generated. A high score indicates that the unknown specimen contains or is associated with the specimen from which the signature was derived.

Alternative signature extraction algorithms can be used as well. One example is the average log ratio method. This algorithm works separately for each group / matrix {A ″ B ″ ... Z ″}.

For each matrix, the signature vector is the row mean of the matrix. Thus, the group {A ″
The signature vector for B ″ ... Z ″} is:

[0078]

[Equation 12]

[Here]

[0080]

[Equation 13]

Is the average of {A ″ ₁₁ , A ″ ₁₂ , ... A ″ _1a }.

Assessment of unknown samples using this approach is performed as follows. Prior to rating, the sample gene expression vector is transformed by taking the log of the ratio in relation to the newly created reference gene expression vector. For example, sample S =

[0083]

[Equation 14]

The conversion of S '=

[0085]

[Equation 15]

Lead to Note that

[0087]

[Equation 16]

Next, a gene having a weak differentiation value is
The remaining rows are removed to be the same as in the signature vector, creating a new sample vector S ″. The score for each signature is then the Euclidean distance between S ″ and the signature vector. Calculated by The normalized score is the standard deviation of (score-randomized score) / randomized score, where the randomized score is the Euclidean distance between S ″ and the signature vector whose gene position is randomized. Is.

The patient data is likewise the database (s) used to carry out the operations described above.
And can be used to improve the algorithm. The database is enhanced by incorporating information about patient sequences or patterns from the discovery database into the analysis database. Thus, increasing the sample size improves the statistical reliability of the matching process (between clinical meaning and sequence). This is true regardless of whether the sequence or pattern represents a negative or positive clinical outcome, provided the results are correct. In addition, some specimens will have sequences or patterns that were not present in the sequences or patterns in the database with which they were compared. These sequences or patterns can provide additional features that will enhance matching when future samples with the same sequence profile are analyzed.

Whether additional confidence can be achieved through the use of additional pattern matching is also considered. That is, different confidence levels can be attributed to matching different patterns. Thus, in order to arrive at a particular diagnosis, the minimum pattern matching may have been substantiated, but the presence or absence of further matching that would be considered redundant under the Diamond model (described below), It can be used to improve confidence in the results.

US Pat. No. 5,692,2 to Diamond
No. 20 proposes a simple set of questions when considering whether to include a given pattern within an algorithm. He first asks what minimal entry data set must exist to establish a positive match with the pattern under consideration. He then asks if there is any single piece of entry data or combination of entry data that excludes or excludes the pattern from further consideration, if any. Finally, he asks if the other patterns already programmed for comparison are hierarchically lower than the pattern being considered. That is, whether other patterns can be "swallowed" by the pattern under consideration.

In the present invention, for the last two questions, part of the process for determining whether and how the algorithm that correlates sequence information with clinical meaning should be modified. The answer is given as. In the Diamond model, a wider pattern will be used if one pattern can only be swallowed by another. However, if additional belief can be achieved by assigning a higher score to data that matches across more patterns, it may be worth retaining the use of both patterns. The same is true when considering whether a single apparently deterministic match should be used rather than the number of pattern matches. The Diamond model only suggests the use of a single match when possible. However, in that case, this may not be desirable if greater statistically significant beliefs can be achieved through the use of many comparison points.

FIG. 1 is a flow chart illustrating a method of incorporating expression profile data within a diagnostic / prognostic algorithm to enhance confidence. Confidence levels, appropriate sample sizes and statistical tools for calculating similar considerations are all well known. Programming the method into executable computer code is likewise conventional and readily accomplished by those skilled in the computer programming arts. The act of performing this process as a continuous and / or pre-programmed process in conjunction with the processing of patient data is one aspect of the inventive method. The example process begins at step 100 with a healthcare provider or other interested party requesting analysis of a patient specimen. At step 200, a sample is obtained and the physical manipulation steps of performing a laboratory assay are performed, either by a health care provider, a laboratory service or a party operating a database system. The culmination of this step is the extraction of the genetic or protein material from which the sequence information is derived. This information is then analyzed in step 300 via comparison with a reference sequence and interrogation via an algorithm. The reference sequence is stored in the analysis database 1000. The algorithms used to perform the analysis may be implemented as part of the programming instructions in the database 1000, or else they are in a separate computer program created to query and manipulate the database 1000. It can be operated via another set of instructions. The analysis in step 300 produces a result (step 310). This result will indicate whether there is sufficient alignment with the reference pattern to provide diagnostic, prognostic or other clinically relevant information. Whether the matching process has identified any previously unidentified patterns, or the identification of previously identified patterns in this sample (or its absence) provides the system with additional statistics. A query is made to determine whether or not (step 32).
0). Additional statistics can be obtained, for example, by increasing the sample size such that increased confidence or prediction power is achieved. The result is step 400
Or, in step 410, they are reported to the party requesting them, or to the location designated as the destination for such results. The results can also be communicated directly to the healthcare provider via electronic communication or any other method. Presenting a pattern that has not previously been identified as of clinical significance, or even more usually the emergence of a pattern previously identified as possibly related to one clinical condition The pattern is tagged if sufficient confidence in the relationship has not yet been substantiated. This tagging is done in step 510. The tagged pattern is stored in the discovery database DB 2000 at step 600. The tag is removed from the data (step 800) upon receipt of a confirmation of the clinical condition (step 700) from the healthcare provider or other person in the position of providing it. The pattern is then moved from the discovery database 200 and into the analysis database 1000 to be used as a reference signature in subsequent analyses. The process can be iterative if multiple patterns are identified, for example by a pattern matching algorithm, and different portions of the pattern correlate different clinical information that requires different confirmation.

The process of the present invention is a standard diagnostic method found in ANN and the prior art (eg clinical chemistry and EIA).
It has the same meaning as that used in (Method) and does not depend on the proof of normal range. In the case of a single or definitive nucleic acid or protein pattern indicative of a disease state or symptom, the presence of any marker (eg gene) has clinical significance. On the other hand, if a combination of markers is used for substantiating a clinical diagnosis, or if statistical beliefs are assigned to one marker group, the pattern compared to the unknown or the sample may change continuously. As long as the pattern can be viewed as "normal", it is a dynamic normal, unlike the normal normally associated with the measured analyte in classical diagnostic medicine. This normal is constantly updated and validated.

Adding patterns from a patient sample into the reference pattern database and database and algorithm of the analysis database presents some challenging problems. For example, how can a previously unseen pattern be known when it can be used to support one diagnosis and weaken the conviction of one diagnosis or to suggest a previously undetermined diagnosis? In the most preferred embodiment of the invention, at the time of the initial analysis, the sequences that are aligned to one database may have some indicia indicating that the diagnosis has not been independently confirmed. "Tagged" with elements). In this most preferred embodiment, the tagged sequence resides in the discovery database. Suppose one sample displays sequences that are consistent with known patterns, and also displays patterns that have not yet been correlated to one disease state or physical condition. Independently, other similar patterns are performed, including a mixture of known patterns and previously unknown patterns. Results based on matches with previously identified patterns are reported, but previously unknown patterns are still not included in the subsequent sample sequence analysis process. The tagged data can be assigned to a data table or database (eg a discovery database). Upon receipt of information confirming a physical condition or disease state and establishing a link between a given clinical condition and a previously unknown pattern, the indicia ("tag") is removed and the sequences are aligned. It can either be fully incorporated into the process, or it can be incorporated into the statistical values that drive the matching algorithm.
Internal registers can be used to attribute the statistical significance to the newly added pattern. That is, a single value may be assigned to the first such "confirmation" of co-occurrence of a pattern and independent confirmation of a disease state, or alternatively, the pattern may be suspect for a given diagnosis. A note may be added to indicate that the information has been broken. If this pattern is seen again and it is correlated with the presence of one disease or condition, it is given a different indicator, eg an indicator that the disease state or physical condition is likely. This process can be followed until the correlation between the presence of the pattern and the disease state or condition is well documented according to well-known statistical methods and criteria.

For the database, this process can be implemented as follows: A large set of characterized patient specimens are treated so that sequences or patterns are identified. For example, about 20 representing two completely different cell or tissue types
A large collection of 0-400 specimens is collected and sequence or pattern data is placed in the discovery database. Bioinformatics methods are used to analyze the discovery database until one pattern is detected that distinguishes two or more different types of cells or tissues when the data is useful. 2. Export the data set needed to define the full pattern range associated with the variable in question to the analytical database. This database is "locked" and used as a clinical reference tool for clinical diagnosis of patients. 3. The diagnosis works by analyzing a new patient with a device designed to measure a predetermined pattern. The new data is compared against an analytical database and a statistical assessment is made of the similarities between the patient sample and the reference pattern. 4. At the same time, the patient pattern is inserted into the discovery database. The new data is combined with all previous data. During each periodic review of the discovery database for new patterns, newly submitted patterns are included in the new dataset. Over time, the statistical value of the discovery set increases and so does the statistical power of the reference pattern. 5. At each time when the reference pattern is derived from the discovery database and they are statistically superior to the preceding pattern, the new pattern replaces the analysis database and acts as the reference pattern.

In the preferred embodiment, the interface between the discovery and analysis databases is "live". In this case, there is no physical separation of the two databases and the analysis domain is defined as a subset within the discovery database. The method of analyzing the discovery database and updating the analysis database reference pattern is continuous.

One important variant to the method is:
This is the case when there are several discovery databases that focus on different patterns. For example, another discovery database can focus on cancers of different organs. Similar to shuffling data from ever-improving discovery databases to analytical databases, it is also possible to merge separate databases to form one large discovery database. The combination of multiple patterns creates an entirely new pattern that is a useful reference for new phenotypes, especially when these patterns are annotated with information about related and unrelated phenotypic features. obtain.

The tagging / tagging process can be accomplished in a number of ways. It is possible to manually perform the tagging and / or tag removal process through the appropriate digitized instructions. For example, when providing information to a recipient of an analysis, the recipient should inform the database operator about the clinical diagnosis once confirmed through a completely different means than genetic testing (eg, biopsy and cell analysis). May be advised. If the requester is in electronic communication with the analysis provider, a simple connection can be created for the requester to enter the verification data directly into the database, thus removing the tag.
Naturally, the situation where the analysis cannot be confirmed must also be considered. In such cases, the tagged data may remain tagged, may be discarded, or may be used to make statistical reports associated with the analysis. Can (for example,
It can be used to reduce confidence in the result). Implementing any of these options is
It is simple from a programming point of view and can be easily achieved by a person skilled in the art.

Preferred Embodiments The method of the present invention can be practiced in many different ways. There are many combinations of sample collection, analysis, reporting, data collection, database and analysis improvement processes. The most preferred combinations are those that match the best capabilities of the various parties involved in the function requiring that capability. In addition, efficiency is another consideration. Given the requirements associated with storing and manipulating large databases with sophisticated algorithms that are constantly being improved in the manner described above, it is most efficient to carry out the analysis process in one or more centralized locations. It is a good thing. This facilitates hardware and software maintenance and upgrade issues and, most importantly, limits the requirements associated with distributing improvements to algorithms and databases. Similarly, sample testing for patterns (ie, the actual laboratory stage) is generally done at the local hospital or reference laboratory because it is the most well-configured and staffed to perform these activities. May be the best.

In the most preferred method, the health care provider obtains the patient specimen in an appropriate format. This will depend on the suspicious disease or condition. For example, when testing for breast cancer, a biopsy specimen of breast cancer may be the appropriate specimen, while a whole blood specimen may be best if the test is general screening. In all cases, the choice of the appropriate sample will be obvious to the parties and will depend on the choice of available assay formats.

After collecting the specimen, the healthcare provider will use the bioinformatics system described herein under appropriate conditions (eg, in a tube containing appropriate preservatives and additives). The specimen is sent to a laboratory that has the ability to obtain the pattern required for the desired analysis. Preferably, but not necessarily, the assay to obtain this pattern is provided by the same party and includes a nucleic acid or protein microassay. Such devices are now well known. Their use is described in US Pat. Nos. 5,143,854; 5,288,644; 5, the disclosure of which is incorporated herein by reference.
324,633; 5,432,049; 5,470,71
0; 5,492,806; 5,503,980; 5,510,
270; 5,525,464; 5,547,839; 5.5
80,732; 5,661,028; 5,848,659;
And numerous patents such as 5,874,219. Preferably the data format is a digital representation of the pattern. This is Gene Expression
Markup Language (GEML ^™ , Rosetta Inpharmatics,
Useful for additional formatting in Kirkland, Washington). This language is an open, documented, open format that allows interchange between gene expression systems, databases and tools. Moreover, this format allows an unlimited number of tags. "Gene E
xpression Markup Langugage (GEML ^™ ), a general data format for exchanging gene expression data and annotations "Rosetta Inpharmatics, www.geml.org/docs/GEM
See L.pdf (2000). This facilitates later confirmation of clinical results and tagging of the data to make the data anonymous, as described below.

The resulting pattern is provided in any form of entry (eg, scanned into a computer that can digitize the pattern) and then analyzed by an operator of the bioinformatics system. The results of the analysis (sequence / pattern matching with the expected diagnosis or condition) are then communicated to the requester. At the same time, the pattern is experimentally maintained in a database associated with the bioinformatics system. Preferably it is tagged as an attempt as described above and kept in the discovery database. The requestor then returns the confirmation information to the operator of the bioinformatics system. If so, then the pattern and any new information that can be found from the pattern becomes part of the analytical database as a reference sequence. The above happens at the same time, in some cases the reception of expression data will confirm the diagnosis of the healthcare provider who has already performed other clinical evaluations. If nothing else was done with the data, the statistical reliability of the analysis would improve through increasing sample size. The database will probably be more robust.

In another preferred embodiment,
The laboratory or health care provider obtains the required specimen. This specimen is assayed by the same tissue as the one performing the analysis. This has some advantages as the assay format and the desired entry format for analysis can be more easily adjusted. At this time, the analysis of the identified patterns and the above-mentioned data / algorithm refinement can be carried out in a similar manner.

Speed up the process in any way that the pattern to be analyzed must be communicated to different locations (for example, one laboratory performing the assay and sending the resulting pattern to a bioinformatics operator). Therefore, it is possible to use electronic communication. Those skilled in the art will readily appreciate Internet and other networked systems for this purpose.

The device of the present invention is best made and used when configured as a specially programmed general purpose computer. In this embodiment, a database system (a combination of a discovery and analysis database combined with programming instructions to function as described above) is a computer or computers specially programmed to perform the functions described herein. Fulfills that function. The instructions may be embodied in any suitable medium for performing computer operations, such as a hard drive, network, optical, or magneto-optical material and others commonly used for this purpose. Articles of manufacture that include media bearing computer instructions for implementing the processes described herein are further embodiments of the present invention.

[Brief description of drawings]

1 is a flow chart illustrating the method of the present invention.

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) G01N 37/00 102 G01N 37/00 102 F term (reference) 2G045 AA25 BA13 BB03 DA12 DA13 DA14 DA36 FB02 JA01 4B029 AA08 AA23 BB20 GB10 4B063 QA20 QQ43 QS25 QS39

Claims (20)

[Claims] 1. A method for providing a clinical diagnostic service: a) collecting a biological specimen; b) analyzing the biological specimen to determine its genetic material composition, its behavior or at least part of its protein. C) reporting the results of the analysis of the biological specimen; and d) incorporating the information obtained through the analysis of the biological specimen into a subsequent analysis of the biological specimen. 2. The method of claim 1, including the step of extracting genetic material from the biological specimen. 3. The method of claim 1, including the step of extracting a protein from the biological specimen. 4. Collection of a biological specimen and extraction of genetic material from said biological specimen is performed by a laboratory service or health care provider, into an analysis and subsequent analysis to determine the composition or behavior of genetic material. The acquisition of information
The method according to claim 2, wherein the method is performed by one business entity that is not a laboratory service or medical provider who performed the collecting and extracting steps. 5. Collection of biological specimens and extraction of proteins from said biological specimens is carried out by laboratory services or medical providers, in order to determine the composition, concentration or behavior of said proteins and in subsequent analyzes. 4. The method according to claim 3, wherein the uptake of such information into the is performed by one business entity that is not the laboratory service or health care provider who performed the collecting and extracting steps. 6. The method of claim 2, further comprising the step of amplifying at least a portion of the genetic material. 7. The method of claim 2, wherein the analyzing step is performed in conjunction with a microarray. 8. A collection and extraction step is performed by a laboratory service or health care provider, and analysis and subsequent incorporation of such information into the analysis to determine the composition or behavior of genetic material comprises collection and extraction. The method of claim 2, wherein the method is performed by one business entity that is not the laboratory service or healthcare provider that performed the step. 9. The collecting and extracting step is performed by a laboratory service or medical provider, and the analyzing and capturing step is performed by one company entity that is not the laboratory service or medical provider who performed the collecting and extracting step. The method of claim 3, wherein the method is performed. 10. The method of claim 1, wherein the analysis is performed by comparing the genetic material, its behavior or the protein with a database containing pattern information. 11. The method of claim 1, wherein the step of incorporating information into a subsequent biological specimen analysis modifies the statistical validity of the results of the analysis. 12. The method of claim 10, wherein the step of incorporating information into the subsequent biological specimen analysis modifies the database. 13. The method of claim 10, wherein the step of incorporating information into a subsequent biological specimen analysis modifies the algorithm used to perform the comparing step. 14. Use of the results of analyzes based on the composition or behavior of genetic material and results not directly based on the composition or behavior of genetic material to determine the probability of the presence or degree of a given physiological condition or disease, The method of claim 1, further comprising the step of performing an additional analysis that is not directly based on the composition or behavior of the genetic material. 15. In a database system for providing clinical diagnosis, prognosis or therapeutic monitoring, comprising a discovery database and an analytical database, the first data entered in the discovery database is the entry of this first data. A database system that modifies the analysis database such that the diagnostic, prognostic or therapeutic monitoring information provided subsequently is given different statistical validity or is analyzed differently than the first data. 16. A machine comprising one or more general purpose computers for performing operations through the database system of claim 15. 17. An article of manufacture comprising a computer-readable medium programmed with one or more components of the database system of claim 15. 18. A method for diagnosing a physiological condition or disease state, comprising: (a) obtaining genetic material from a subject; (b) determining an expression pattern of the genetic material; (c) a discovery database and an analytical database. By using a database system for providing clinical diagnosis, prognosis or therapeutic monitoring, including
Correlating the expression pattern with a physiological condition or disease state; and (d) incorporating information about the genetic material in the database so as to modify the database. 19. (e) A normal specimen from normal tissue and diseased human tissue to produce an analysis of normal reference genes from normal human tissue and diseased reference genes from diseased tissue. Performing steps (a) to (d) on the affected sample, (f) storing the normal reference gene analysis and the affected reference gene copy image analysis in a database, (g) obtaining a subject sample from a subject And generating a genetic analysis by performing steps (a)-(d) from the subject's sample; and (h) identifying at least one of the reference analyzes that approximate the patient sample based on the database. To do so, further comprising the step of processing the genetic analysis of the subject specimen with an algorithm driven device.
The method according to 8. 20. The method of claim 18, wherein step (d) is performed continuously.